Hybrid Branch Predictor Having Negative Ovedrride Signals

1. A processor comprising: a plurality of three or more branch predictors, wherein each branch predictor is adapted to provide a prediction and an override signal, each override signal being produced using an independent override logic associated with each branch predictor; and a branch prediction control circuit to generate a branch prediction based on the prediction and override signal from each predictor, the branch prediction control circuit being operable to: determine if a negative override signal has been asserted from one of the plurality of three or more branch predictors; if a negative override signal has been asserted from one of the plurality of three or more branch predictors, then select a branch prediction from a predictor having a highest confidence level that did not assert the negative override; and otherwise, if a negative override signal has not been asserted from one of the plurality of three or more branch predictors, then select a branch prediction from a predictor having a highest confidence level from among the plurality of three or more branch predictors.

2. The processor of claim 1 wherein each branch predictor is adapted to provide a prediction, an override signal and a confidence level.

3. The processor of claim 1 wherein each branch predictor is adapted to provide a prediction and an override signal, the override signal to include a positive override signal to indicate when the predictor asserting the positive override signal should be trusted.

4. The processor of claim 1 wherein each branch predictor is adapted to provide a prediction and an override signal, the override signal to include a negative override signal to indicate when the predictor asserting the negative override signal should not be trusted.

5. The processor of claim 1 wherein the override signal provided by at least one branch predictor of the plurality of three or more branch predictors comprises: an override signal output by a corresponding branch predictor of the plurality of three or more branch predictors, the override signal configured to result in either a selection or a non-selection of the prediction of the corresponding branch predictor regardless of a short term history of the plurality of three or more branch predictors.

6. The processor of claim 1 wherein each override signal provided by each branch predictor of the plurality of three or more branch predictors comprises an independent override signal produced using data kept local to each branch predictor.

7. A method comprising: determining if a negative override signal has been asserted from one of a plurality of three or more predictors, each override signal being produced using an independent override logic associated with each branch predictor; if a negative override signal is determined as being asserted from one of the predictors, then selecting a branch prediction from a branch predictor having a highest confidence level that did not assert the negative override signal; and otherwise, if a negative override signal is not determined as being asserted, then selecting a branch prediction from a branch predictor having a highest confidence level from among the plurality of three or more predictors.

8. A method comprising: providing a plurality of three or more override signals from a corresponding plurality of predictors, each override signal being produced using an independent override logic associated with each predictor; determining if a negative override signal has been asserted from one of the plurality of predictors; if a negative override signal has been asserted from one of the plurality of predictors, then selecting a branch prediction from a predictor having a highest confidence level that did not assert the negative override; and otherwise, if a negative override signal has not been asserted from one of the plurality of predictors, then selecting a branch prediction from a predictor having a highest confidence level.

9. A branch prediction system, comprising: determination means for determining if a negative override signal has been asserted from one of a plurality of three or more predictors; and selection means for selecting: a branch prediction from a predictor having a highest confidence level that did not assert the negative override signal if the negative override signal has been asserted from one of the plurality of three or more predictors, or a branch prediction from a predictor having a highest confidence level if the negative override signal has not been asserted from one of the plurality of three or more predictors.

10. A branch prediction system, comprising: prediction means for providing a plurality of three or more override signals from a corresponding plurality of predictors, each override signal being produced using an independent override logic associated with each predictor; determination means for determining if a negative override signal has been asserted from one of a plurality of predictors; and selection means for selecting: a branch prediction from a predictor having a highest confidence level that did not assert the negative override signal if the negative override signal has been asserted from one of the plurality of three or more predictors, or a branch prediction from a predictor having a highest confidence level if the negative override signal has not been asserted from one of the plurality of three or more predictors.